Broadwell to Skylake, will be similar to the transition from Prescott to Conroe

[crashtech]

So you don't think an i5-2500k will be playing new games till 2020? Considering its been almost 4 years since it's launch and still no single gpu card can bottleneck it.

That's unknown, but I was addressing the assertion that an X2 4600+ is suitable for games today, when it's really not unless you stick to old titles.

 

[Dave2150]

I dont really see the lower power consumption, since desktop Skylake quad core has a 95 watt TDP, even higher than Haswell.

This may simply be a 4.5-4.8Ghz clockspeed CPU - in order to entice sandy bridge owners to upgrade etc.

I'm sure Intel could have released a 95watt mainsteam Haswell 4795 type CPU if they really wanted to, clocked slightly higher than the 4790k.

Slightly more voltage = 7 more watts of TDP.

 

[B-Riz]

Do you have any idea how old those games are?

I think you need a reality check, it's 2015 and there are many new games released this year and last year that have no hopes in hell of running on your CPU.

2500K is a bottleneck for many GPU's already, when running in Crossfire/SLI.

Even a single TitanX is bottlenecked by the 2500k, unless you're one of the 1% who clock it to 5Ghz+

Is the 2500k with a nice OC really going to bottleneck in CF / SLI?

Anyone here with that setup care to chime in?

Anyone have benchmarks to reference?

My Google-fu is busted right now, cannot find yay / nay benchmarks.

Found this: http://www.ocaholic.ch/modules/smartsection/item.php?itemid=1158&lang=english

In 4K, it may not be ideal, but 1080p and 1440p should still be fine with it...

 

[AtenRa]

Even a single TitanX is bottlenecked by the 2500k, unless you're one of the 1% who clock it to 5Ghz+

The vast majority of Core i5 2500K were capable of 4.3-4.5GHz without the need of extreme cooling or very high voltage.

At 4.5GHz those SBs are close to 90-95% performance or more of current Haswell 4GHz at 1080/1440p or higher resolutions (high Quality Image settings and AA filters).

With DX-12 Games, those SBs will be more than enough even with next year 14nm high-end GPUs.

 

[*NixUser]

Broadwell to Skylake, will be similar to the transition from Prescott to Conroe

What is the possibility of this being true?

 

[cytg111]

1%

but then again, those israeli values for 1% can be pretty big.

 

[Dave2150]

Is the 2500k with a nice OC really going to bottleneck in CF / SLI?

Anyone here with that setup care to chime in?

Anyone have benchmarks to reference?

My Google-fu is busted right now, cannot find yay / nay benchmarks.

Found this: http://www.ocaholic.ch/modules/smartsection/item.php?itemid=1158&lang=english

In 4K, it may not be ideal, but 1080p and 1440p should still be fine with it...

Comparing Titan X users benchmark and games results, theres a huge difference depending on the CPU used and the speed that it's clocked at.

I havent found a professional review testing TitanX and TitanX SLI yet, that would be interesting to confirm that they are CPU bottlenecked.

Heres one professional review using 780TI's in SLI:

http://www.guru3d.com/articles_pages/core_i7_5960x_5930k_and_5820k_processor_review,17.html

Obviously the difference between 780ti SLI and Titan X SLI is going to be huge, especially when the resolution is cranked up to 4K.

Those on X58 or sandy bridge won't get the most out of the fastest GPU's available, that's for sure.

 

[MrTeal]

Comparing Titan X users benchmark and games results, theres a huge difference depending on the CPU used and the speed that it's clocked at.

I havent found a professional review testing TitanX and TitanX SLI yet, that would be interesting to confirm that they are CPU bottlenecked.

Heres one professional review using 780TI's in SLI:

http://www.guru3d.com/articles_pages/core_i7_5960x_5930k_and_5820k_processor_review,17.html

Obviously the difference between 780ti SLI and Titan X SLI is going to be huge, especially when the resolution is cranked up to 4K.

Those on X58 or sandy bridge won't get the most out of the fastest GPU's available, that's for sure.

Wouldn't the opposite be true, where at 4k the differences between CPUs would drop as the framerate tanks? Titan Xframerates @4k will be lower than 780Ti at the 1080P resolution that test used.

 

[mohit9206]

What is the possibility of this being true?

As much as John Cena's possibility of turning heel.

 

[Phynaz]

What is the possibility of this being true?

0%

 

[frozentundra123456]

Depends on how loosely you define "similar". After all wasnt the source WCCF Tech? They are never wrong.

Edit: Seriously though, I was hoping for a decent improvement from Haswell, especially since it is 2 generations. But the longer we go without any leaks of performance, the less optimistic I become.

 

[Dave2150]

Depends on how loosely you define "similar". After all wasnt the source WCCF Tech? They are never wrong.

Edit: Seriously though, I was hoping for a decent improvement from Haswell, especially since it is 2 generations. But the longer we go without any leaks of performance, the less optimistic I become.

Yeh I agree with you. Surely we'd have seen some leaks or hints from reputable sources of the huge performance increase, if there were one.

There's no way that those with engineering samples wouldn't leak a benchmark of two if these really were something special.

I'm not convinced that Skylake will even release for Desktop this year, the way things are going.

Sales of Haswell 22nm Desktop chips are still soaring - whilst the amount of

 

[ocre]

I dont really see the lower power consumption, since desktop Skylake quad core has a 95 watt TDP, even higher than Haswell.

Yeah, there is a scary trend here. Intel gained very little in the high end this past node shrink. 22nm ivy at over 4ghz used as much or more power than 32nm sandy. There has been a real issue here that most people havent caught on to. This picture is worth 1000 words.

The only improvements to TDP has been in lower frequencies. If Haswell hadnt improved ipc, we would have gained absolutely nothing these past 2 generations. The ipc gains were rather small but they were at least something.

But the problem is really disturbing, the 4790k came out on a very very mature 22nm silicon. Intel claims only a tiny 4 watt increase for the 4790k but in the real world it is much more.

I am not trying to make the point that there was no improvement going from 32nm to 22nm, i dont want people to get carried away with that. Lets just say there was very very little improvement when it comes to the higher speeds and power consumption.

My fear is that broadwell has ran into the same troubles. I think it is with Intel's 14nm node. This is a scary thing to think about. It really is gonna take a radical new architecture to move us forward. Skylake might be this awesome tech, but if they have to run it at 3ghz........
its gonna really hurt. Loosing clock speed will be a negative that takes away from how great the architecture could be.

We have hit a major cliff and intel was the first to come on it. Now we see TSMC finding 20nm HP to be pretty much worthless. Everyone is now thinking it is just 20nm. People are thinking all will be fine when tsmc goes to 16nm but i am not so sure now.
When i see something funny going on with intel and their smaller nodes, then TSMC and their 20nm HP. I really dont like the way things are looking right now

 

[Dave2150]

Yeah, there is a scary trend here. Intel gained very little in the high end this past node shrink. 22nm ivy at over 4ghz used as much or more power than 32nm sandy. There has been a real issue here that most people havent caught on to. This picture is worth 1000 words.

The only improvements to TDP has been in lower frequencies. If Haswell hadnt improved ipc, we would have gained absolutely nothing these past 2 generations. The ipc gains were rather small but they were at least something.

But the problem is really disturbing, the 4790k came out on a very very mature 22nm silicon. Intel claims only a tiny 4 watt increase for the 4790k but in the real world it is much more.

I am not trying to make the point that there was no improvement going from 32nm to 22nm, i dont want people to get carried away with that. Lets just say there was very very little improvement when it comes to the higher speeds and power consumption.

My fear is that broadwell has ran into the same troubles. I think it is with Intel's 14nm node. This is a scary thing to think about. It really is gonna take a radical new architecture to move us forward. Skylake might be this awesome tech, but if they have to run it at 3ghz........
its gonna really hurt. Loosing clock speed will be a negative that takes away from how great the architecture could be.

We have hit a major cliff and intel was the first to come on it. Now we see TSMC finding 20nm HP to be pretty much worthless. Everyone is now thinking it is just 20nm. People are thinking all will be fine when tsmc goes to 16nm but i am not so sure now.
When i see something funny going on with intel and their smaller nodes, then TSMC and their 20nm HP. I really dont like the way things are looking right now

Haswell greatly improved idle or low usage power consumption over Ivy Bridge.

For the vast majority of users out there, their PC is running at idle, or low CPU utilization for a significant portion of the day.

 

[ocre]

Yep,
Haswell was the 22nm tock. The architectural improvement. Ivy was the tick. Intel put a lot of effort in driving down consumption which we see results in the low end of the spectrum. I am not suggesting that intel cannot improve their architectures, not that at all.

My concerns were the nodes. Node shrinks not giving us much in the high end. We used to get higher mhz as well as lower consumption, all as a result of the node shrink alone. But their 22nm brought us very very little compared to 32nm or 45nm. Very very little in the high end, it didnt raise the bar on overclocking nor consumption at high mhz.
I fear that 14nm is gonna follow suit.

Skylake will be the 14nm tock. It will be the architectural improvement. We wont even see much at all of the 14nm tick. Think about it. That is really a big deal.

There is no doubt, intel can improve their architecture. But with a tick tock strategy, the path forward seemed very sound. But now there are two ways to look at it, either we are inching forward dragging a limp leg or we are driving on two flats. Either way, things are not moving forward as planned. Now we have to completely depend on the architectural improvements of skylake, it all rest on it. But if intel has to reduce the clock speed because of their 14nm silicon, it hurts even worse.

Everything now rest on skylake. I believe not only is 14nm problematic for moving up in the higher frequencies, we might even have to move down some from the 4790k/4770k. So whatever architectural improvements skylake might give us, it might be limited by the node. This is devastating if your path forward is tick (node advancement)then tock (architectural advancement)

 

[Idontcare]

I fear that 14nm is gonna follow suit.

What is to fear?

Keep your 32 or 22nm CPU, your 256 or 512GB SSD, and spend your upgrade money on the things that are still rapidly changing while giving you something you value...like a 4K monitor (or two) and juiced up GPUs to go with it!

The only people who have anything to "fear" are Intel employees and Intel shareholders. And while I know plenty of both, but am neither myself, I feel like they kinda made their bed and they deserve to lay in it for a while.

Personally I'm going with 4K monitors (yes plural) as my "2015" upgrade de l'année.

 

[Qwertilot]

Or just get one of the 35/45 variants and live content with the more than enough performance those now give

 

[frozentundra123456]

If the leaks are correct, Broadwell K already is over 3ghz at 65 watts, so we may be getting a little hysterical here thinking skylake will be a regression to only 3 ghz or less. I agree the days of big gains from die shrinks are over, but like IDC kind of said, we pretty much have all the performance we need on the desktop anyway.

 

[frozentundra123456]

What is to fear?

Keep your 32 or 22nm CPU, your 256 or 512GB SSD, and spend your upgrade money on the things that are still rapidly changing while giving you something you value...like a 4K monitor (or two) and juiced up GPUs to go with it!

The only people who have anything to "fear" are Intel employees and Intel shareholders. And while I know plenty of both, but am neither myself, I feel like they kinda made their bed and they deserve to lay in it for a while.

Personally I'm going with 4K monitors (yes plural) as my "2015" upgrade de l'année.

Not sure you meant it that way, but i dont really know what intel employees have done to deserve negative consequences, except for top management perhaps.

 

[csbin]

Intel Confirms Skylake Launch In 2H 2015 Around IDF15 – 100-Series Chipsets Shipping in May 2015


http://wccftech.com/intel-confirms-skylake-launch-2h-2015-idf15-100series-chipsets-shipping-2015/

 

[Idontcare]

Not sure you meant it that way, but i dont really know what intel employees have done to deserve negative consequences, except for top management perhaps.

We all choose our career paths, hitch our wagons so to speak. Choose poorly and who should we blame but ourselves?

 

[frozentundra123456]

We all choose our career paths, hitch our wagons so to speak. Choose poorly and who should we blame but ourselves?

Maybe someone with your qualifications can pick and choose the job they want, but a lot of people cant. Many people are working for intel and AMD because that is the best (or only) position they could find. Normally I agree with most of your posts, but I think you are being a bit harsh in this case. Personally, I searched for a new job in my field for almost one year before finally finding one just before my previous job ran out, so I know how hard it can be.

 

[therealnickdanger]

Intel Confirms Skylake Launch In 2H 2015 Around IDF15 – 100-Series Chipsets Shipping in May 2015

ES benchmarks shouldn't be too far away.

 

[ocre]

Not sure you meant it that way, but i dont really know what intel employees have done to deserve negative consequences, except for top management perhaps.

I think he meant intel as a whole. They chose to chase after the ultra low power chips and these nodes will only benefit the low end. I find it sad, as a fan of PCs for so many years of my life.

It is like, Intel the massive lion got distracted chasing mice and missed the large elk standing by. The ultra low power ARM chips are very low margin and exist in an ecosystem that intel has struggled and struggled to make progress in. Yet they kept on putting all their focus there, at all cost. the upper segment of high performance has suffered as a result. And by the looks of it, with 14nm it might even drop off.

What an incredibly dumb path for them to take. They are completely and totally vested in this and you cant turn such a big move around. Arm chips are racing upwards in performance in the meantime........

So, think about it. I am pretty sure this is how he sees it

 

[Enigmoid]

It is like, Intel the massive lion got distracted chasing mice and missed the large elk standing by. The ultra low power ARM chips are very low margin and exist in an ecosystem that intel has struggled and struggled to make progress in. Yet they kept on putting all their focus there, at all cost. the upper segment of high performance has suffered as a result. And by the looks of it, with 14nm it might even drop off.

I agree. Intel sells its chips and bases its reputation on its performance (its brand). Once they start losing the performance crown you will start to see them struggle.